Background semi-structured interviews

The interviews for this study have been conducted between May 18^th 2021 and June 8^th . Because of convenience and the Covid-19 pandemic, the interviews have been conducted online via Microsoft Teams. Interview candidates and organizations have been contacted via e-mail, LinkedIn and via connections within the researcher’s direct network. The interview candidates need to have experience in managing an AI transition or implementing an AI application in the organization. The target for the amount of interviews was a minimum of 5 and maximum of 10 interviews. However, only an interview count of 4 has been achieved due to a low response rate and time constraints.

Qualitative study approach

The semi-structured interviews were conducted online via Microsoft Teams, and recorded (with verbal consent) with a recording device. During the interview, the researcher made notes on paper of terms that potentially could be used for coding and gave a quick overview

32

of the topics discussed during the interviews. The interviews have been transcribed by using an AI leveraged audio translation site, Otter.ai, this software transcribed the audio verbatim.

The transcription of the software was quite accurate, the transcription got read and used as input for the coding process. The interview analysis followed a mixed approach of deductive and inductive coding as described by Babbie (2013). Before carrying out the interview a deductive approach was taken to identify themes likely to be discussed based on the

systematic literature review, based on this (deductive) codes have been created. This helps the researcher ask more pointed questions during the semi-structured interview. After the

interviews had been conducted, an inductive approach has been taken to observe the

frequency of words, patterns and contexts mentioned in the interview to further extend and improve the codebook. The open coding method (Babbie, 2013) has been used to generate coding concepts and categories. This in turn, has been used to identify practices and methods used to conduct an AI transition in business processes and answer KQ.1.

After the open coding process is done, the researcher makes use of axial coding (Babbie, 2013) to classify the codes identified in broader categories and determine relations between them. This helps to get a more comprehensive view and helps to determine possible

relationships between categories and codes which might get lost when just open coding is done. The coding scheme is presented in Figure 4.2. below.

Figure 4.2. Coding scheme

33 Content of the semi-structured interviews

Before starting the interview, the interview protocol (Section A.2.), information sheet for semi-structured interview (Section A.3.) and an informed consent form (Section A.4.) have been sent via e-mail. This is done to give full disclosure to the interviewee about the content of the interview, the ethical implications and the rights of the interviewee. The interview protocol gives a framework of the main questions that will be asked, because it is a semi structured interview, the interview is more loosely structured. This allows for the interviewees to express themselves more freely and gives the interview more of a conversation-like flow.

Ethical Implications of the semi-structured interviews

As mentioned, an information sheet and an informed consent form have been sent to the interviewees before the interview. The information sheet contains information on the purpose of the research, benefits and risks of participating, procedures for withdrawal from the study and privacy and personal data. After the interviewee has read and understood the content of the information sheet, the interviewee agrees on the content of the informed consent sheet.

The informed consent form contains seven questions verifying that the interviewee knows the implications of taking part in the study and how the information gained from the interviewee will be used in the study. The interviewee has the choice to send the consent form physically to the researcher or consent verbally on record.

The content of the information sheet and the informed consent form have been approved by the Ethics Committee of the Faculty of Behavioral, Management and Social Sciences at the University of Twente.

Description semi-structured interview respondents

For this interview, four respondents that fulfilled the requirement of having experience in managing an AI transition or implementing an AI applications in an organization have been found. The interview respondents are diverse, two from consulting, one from the service industry and one from a governmental organization. Furthermore, the respondents vary in age and professional experience. This gives, even with the low response rate, a broad vision of the viewpoints of AI practitioners. Find below in Table 4.2. an overview.

Table 4.2. Description interviewees

Code Function Experience Organization Pseudonym

IR-1 Senior consultant 5 years Manufacturing Consulting

IR-2 Junior consultant 2 years IT Consulting

IR-3 Chief Information Officer (CIO)

20+ years Service industry

IR-4 IT manager 10+ years Governmental organization

Main topics semi-structured interview

The main topics the interview focuses on are the following: Meaning of Artificial

Intelligence, Management Techniques for AI transition, AI transition KPIs, Challenges to managing AI transition, Best practices managing AI transition and Future of AI according to practitioners. These topics give a broad insight into the current state of AI transition

management according to practitioners. Besides, these interview topics touch on similar topics as in the systematic literature review. This helps to give a multi-faceted perspective on the topic in question. Find below in Table 4.3. an overview of the insights gathered on

34

management techniques to conduct an AI transition according to the Interview Respondents (IRs).

Table 4.3. Management techniques AI transition interviewees

Management technique Interviewee

Agile Methodology IR-1, IR-2 and IR-4

Design Thinking combined with customer journey IR-2

In Table 4.4. the requirements of executing the Vision as mentioned by the interviewees are presented.

Table 4.4. Vision execution requirements interviewees

Vision requirement Interviewee

Leadership – Have a business sponsor IR-3, IR-4

Investment – Need support from top management IR-2, IR-3,

IR-4 Education – Employees interacting with AI need basic AI capabilities IR-2, IR-3,

IR-4

Talent – hire right talent to execute vision IR-1, IR-4

Compliance – Privacy, transparency and protection IR-1, IR-2, IR-4 Key Performance Indicators (KPIs) – Set clear KPIs for executing the vision IR-1, IR-2,

IR-3

Culture – Forging Human-AI symbiosis IR-3

In table 4.5. the vision requirements and management techniques identified (in italics) are categorized based on the identified categories in Table 4.1. of Section 4.1. The category

‘metric specification’ has been changed to ‘continuous feedback’, because metric

specification is an important aspect of ensuring continuous feedback as mentioned in the Agile methodology and KPIs sections of this chapter.

Table 4.5 Categorized practices and vision requirements for AI implementation

Identified categories Identified Practices and vision requirements AI vision Define solid AI planning and vision

Design an organizational structure which is supportive for AI Design Thinking combined with customer journey

AI vision identification

Culture – Forging Human-AI symbiosis Process identification Define processes clearly

Build an integrated AI system Continuous feedback Define metrics clearly

Agile Methodology

Key Performance Indicators (KPIs) – Set clear KPIs for executing the vision AI leadership Establish an AI leader

Recruit AI talent and train AI capabilities current staff Leadership – Have a business sponsor

Talent – hire right talent to execute vision

Investment – Need support from top management

35 AI governance Define AI governance

Initiate and maintain foundational data capabilities Education – Employees need basic AI capabilities Compliance – Privacy, transparency and protection